Like Evan, I've failed to reproduce this, even with amd64, and even with dding /dev/urandom over the start of the region where the swap partition is going to live first. But maybe I was just unlucky.
My best guess is still that libparted's swap initialisation code might be at fault. I'm going to try dropping that in favour of mkswap. If it recurs, please reopen this bug.
Like Evan, I've failed to reproduce this, even with amd64, and even with dding /dev/urandom over the start of the region where the swap partition is going to live first. But maybe I was just unlucky.
My best guess is still that libparted's swap initialisation code might be at fault. I'm going to try dropping that in favour of mkswap. If it recurs, please reopen this bug.